Interesting issue, but this seems like the worst possible solution.

Voting for the electorate candidate who is going to come a distant third might be a nice gesture, but has the same effect as abstaining. Most minor party supporters still have a preference between the two main electorate contenders, which is why they use their electorate votes "strategically" (personally I think calling this "strategic" voting is overstating it; “pragmatic” or “effective” voting would be better terms). Given that electorate contests are generally a two-horse race, it makes sense to back whichever of the real contenders you prefer (or dislike the least) rather than your ideal first preference third-place candidate (indicating your true party preference is what the party vote is for).

Incidentally, I think the approach Kevin Hague took here is pragmatic and entirely appropriate (and not because I am a rabid Labour supporter, but because it is an example of voting sensibly): http://www.3news.co.nz/Green-candidate-votes-for-Labour-opponent/tabid/209/articleID/233950/Default.aspx

The suggested "best loser" system would change these incentives, and likely mean that areas with strong Green support end up being represented by National MPs, and that any areas with strong ACT or Conservative support end up being represented by Labour MPs (hardly appropriate representation, surely?). The system would also heavily favour minor party candidates running in safe seats (for either of the main parties), where the electorate vote is less important. On that basis, I would strongly oppose such a system, at least in the absence of preferential voting.